Closed SpursLipu closed 3 years ago
I got the same issue.I only found ratio=0.4 in the original paper,and I can't get the best result either.
Already mentioned in #8. If you cannot find it: (ratio=0.4, d1=24, d2=33, rotate=1, prob=0.8, mode=1, epochs=280)
I find your work mention GridMask achieve great preformance on cifar 10 dataset but I can't get this result. Could you tell me some hyper-parameters in cifar10 dataset like d1, d2, rotate and ratio?